For taskassigning project，I want the same tasktype tasks can be adjacent as more as possible。 - optaplanner

Question

Below is what I plan, please give me advice:
(1) In the Task class, add a CustomShadowVariable:
#CustomShadowVariable(variableListenerClass =
NumOfAdjacentUpdatingVariableListener.class,
sources = { #PlanningVariableReference(variableName =
"previousTaskOrEmployee") })
private Integer numOfAdjacent;
(2) In the class NumOfAdjacentUpdatingVariableListener:
#Override
public void afterEntityAdded(ScoreDirector<TaskAssigningSolution>
scoreDirector, Task task) {
//update number of adjacent tasks : numOfAdjacent
}
#Override
public void afterVariableChanged(ScoreDirector<TaskAssigningSolution>
scoreDirector, Task task) {
//update number of adjacent tasks : numOfAdjacent
}
(3) In the constraints，add a constraint to reward the numOfAdjacent.

Answer 1

You can achieve that without a custom shadow variable. Simply add a new constraint that penalizes every pair of adjacent tasks if they have different types. This constraint will push OptaPlanner to reorder tasks and minimize the number of adjacent tasks with different types.
The constraint could look like this:
private Constraint adjacentTaskDifferentTypes(ConstraintFactory constraintFactory) {
return constraintFactory.forEach(Task.class)
.join(Task.class, equal(Task::getNextTask, Function.identity()))
.filter((t1, t2) -> !t1.getTaskType().equals(t2.getTaskType()))
.penalize("Adjacent task have different types",
// constraint weight
BendableScore.ofSoft(HARD_LEVELS, SOFT_LEVELS, 0, 3));
}
This constraint works on the TaskAssignment example in https://github.com/kiegroup/optaplanner. Try implementing a similar constraint using your project's domain model.
Then tweak the resulting solution by changing the constraint's priority. You can adjust the priority by changing the constraint weight:
Change the match weight (3 in the code snippet above) to change the constraint priority relative to other constraints affecting the same score level.
Change the constraint's score level (0 in the code snippet above) to give the constraint higher priority than all constraints on lower score levels.